摘要
运用带动量因子的最速下降BP神经网络实现对刀具磨损状态识别。首先标定相机并采集刀具图像;经过倾斜校正、边缘定位、坐标变换和曲线拟合等技术处理后,提取刀尖后角、刀尖夹角、刀尖半径和刀尖高度4种特征参数并融合构成特征矢量组;最后建立改进BP神经网络模型,对刀具磨损情况进行训练和预测。仿真实验证明该方法具有很好的刀具磨损识别准确度和检测效率。
The realization of the tool wear state recognition by using BP neural network based on the steepest descent Algorithm with momentum factor.First,camera is calibrated and tool wear images are collected.After some techniques processes such as tilt correction,edge location,coordinate transformation and curve fitting,acquirement of 4 kinds of characteristic parameters including blade angle,tip angle,nose radius and tip height.A feature vector group is constituted with combination of parameters.Finally the improved BP neural network model is established to train and to predict the tool wear.
出处
《工业控制计算机》
2015年第10期49-51,53,共4页
Industrial Control Computer
基金
人工智能四川省重点实验室资助项目(2012RZY22)
机械装备智能信息工程学科校级交叉学科建设项目(2014JC02)
四川理工学院2014年研究生创新基金项目(y2014006)
关键词
改进BP神经网络
刀具磨损
图像处理
参数融合
状态识别
improved BP neural networks
tool wear
image processing
parameter integration
state recognition
